What AI Overviews do to your traffic, and how to see it
Impressions hold, positions hold, clicks fall, and nothing you changed explains it.
The signature
Position stable, impressions stable or rising, clicks falling. An impression now includes results where a synthesised answer sat above everything and satisfied the user before they reached you.
It hits informational queries first
Queries phrased as questions are the most summarisable and lose the most. Queries with an implied action, where the user wants to do something rather than know something, are far more resistant.
What to do about it
Shift weight toward queries where the answer is a destination rather than a paragraph. This is the same conclusion as the blog-versus-tool measurement, arrived at from a different direction, which is usually a sign it is right.
Measure it on your own queries
The industry-wide numbers are averages across query types that behave nothing alike, so they cannot tell you anything about your site. Take your top twenty queries, search each one, and record whether an AI answer appears and whether it makes your result redundant. Twenty minutes of that beats every published study, because it is about you.
The pages most exposed
Definitional and single-fact pages, which is most of what gets written as introductory content. Least exposed: anything requiring an action, a live calculation, a login, or a judgement about a specific case. On this site the exposed category and the low-performing category were largely the same pages, which made the strategic response easy: it was already the right thing to do.
Measured, not asserted
An impression is worth less than it was, which is why this site's CTR fell from 6.02% to 2.72% across three months while clicks rose from 4,409 to 16,410. Both facts are true at once, and they have the same cause.
Free tool for this: Noise or Signal. No account, nothing uploaded.
Where this goes deeper
Every number on this page comes from one complete dataset: one product taken from zero to 100K+ users on search alone, with nothing spent on advertising. The full argument is Chapter 11 and Chapter 16 of the book. Five chapters are free to read.